People with diabetes face a life-long optimization problem: to maintain strict glycemic control without increasing their risk for hypoglycemia [13, 58, 59]. The engineering challenge related to this problem is to design algorithms using automated insulin delivery to exert optimal closed-loop control of glucose fluctuations. Since the early studies of continuous external glucose regulation (e.g., BioStator, [10]), two primary approaches have emerged: The use of classic proportional-Integral-derivative (PID) algorithms, and modern methods based on models of the human metabolism. The first studies using subcutaneous insulin delivery and continuous glucose monitoring (CGM) employed PID control [57,60]. Recently, model predictive control (MPC), received considerable attention [20, 21, 44, 50] due to its many clinical and engineering advantages.